Calibration and Validation

Calibration and Validation

This page introduces the calibration and validation of OSMOSE configurations. Calibration is the process of estimating model parameters so that simulated outputs are consistent with available observations, while validation focuses on evaluating whether the calibrated model reproduces patterns or indicators that were not directly used for fitting.

OSMOSE calibration commonly relies on the calibrar package, which provides tools for parameter estimation in complex models. Calibration workflows can include biomass indices, catches, size distributions, survey outputs, fishing information, and other data sources depending on the model application and available observations.

The pages in this topic describe the practical requirements for running a calibration, the construction of objective functions, optimisation strategies, diagnostics, and troubleshooting. Some examples were originally developed in the context of the interannual calibration of an OSMOSE configuration, such as OSMOSE Peru, but the general principles apply to other OSMOSE applications.

Pages in this topic

• Installation
  Describes the packages and software requirements needed to run OSMOSE calibration workflows.

• Overview
  Introduces the general calibration workflow, including model initialisation before calibration and the R tools used to initialise the calibration itself.

• Using calibrar
  Describes the general functioning of calibrar and how it is used to support OSMOSE parameter estimation.

• Data requirements
  Summarises the types of data that can be used for calibration and validation, including abundance, biomass, catch, size-structured data, survey indicators, and auxiliary information.

• Objective functions
  Describes how model-data differences are formalised in objective functions, including likelihood-based or penalised approaches.

• Optimisation
  Describes optimisation methods and practical considerations for estimating OSMOSE parameters.

• Multi-phase calibration
  Explains how calibration can be organised into successive phases, from simpler or less dependent parameters to more complex and time-varying parameters.

• Diagnostics
  Describes checks used to evaluate calibration results, model fit, parameter behaviour, convergence, and consistency with observations.

• Troubleshooting
  Lists common calibration issues and possible solutions.

How to use this page

Users preparing a first calibration should start with Installation, Overview, and Using calibrar.

Users defining a calibration setup should then read Data requirements, Objective functions, and Optimisation.

Users working on complex applications should read Multi-phase calibration and use Diagnostics and Troubleshooting to review results.

For outputs used in calibration and validation, see Outputs and Observation. For practical calibration examples, see Calibrate a model.